Rey et al. 2020 data analysis

Resources

• https://deneflab.github.io/MicrobeMiseq/demos/mothur_2_phyloseq.html
• https://joey711.github.io/phyloseq/plot_ordination-examples.html
• https://david-barnett.github.io/microViz/articles/web-only/ordination.html
• https://www.gdc-docs.ethz.ch/MDA/handouts/MDA20_PhyloseqFormation_Mahendra_Mariadassou.pdf
• http://rstudio-pubs-static.s3.amazonaws.com/266780_cac4994322494658904507a7606b1dd8.html

Running this notebook

To run run this notebook in its entirety, clone or download the iobis/pacman-pipeline-training repository. The notebook can be found in the datasets/rey/analysis directory.

Importing data

The PacMAN pipeline exports results as a phyloseq object which can be imported into R for analysis. This is how phyloseq objects are structured:

Figure from https://joey711.github.io/phyloseq/import-data.html

Read the phyloseq object and verify that we have a OTU table, a sample table, and a taxonomy table. A phylogenetic tree has not been calculated so that slot is empty.

physeq <- readRDS("../results_noblast/phyloseq_object.rds")

physeq

## phyloseq-class experiment-level object
## otu_table()   OTU Table:         [ 5327 taxa and 22 samples ]
## sample_data() Sample Data:       [ 22 samples by 20 sample variables ]
## tax_table()   Taxonomy Table:    [ 5327 taxa by 14 taxonomic ranks ]

While we can access these tables individually (for example using physeq@sam_data), the phyloseq package also has a function psmelt() to convert the phyloseq object into a single large data frame:

library(dplyr)

df <- psmelt(physeq) %>%
  # convert empty strings to NA across all character columns
  mutate_if(is.character, ~na_if(., "")) %>%
  # convert to tibble for prettier printing
  as_tibble()

df

## # A tibble: 117,194 × 37
##    OTU    Sample Abund…¹ eventID mater…² event…³ verba…⁴ local…⁵ verba…⁶ event…⁷
##    <chr>  <chr>    <int> <chr>   <chr>   <chr>   <chr>   <chr>   <chr>   <chr>  
##  1 asv.2  SAMN1…   25147 SAMN10… SAMN10… settle… 43.28 … Spain:… 7m      2017-1…
##  2 asv.7  SAMN1…   16275 SAMN10… SAMN10… settle… 43.28 … Spain:… 7m      2017-0…
##  3 asv.8  SAMN1…   15862 SAMN10… SAMN10… settle… 43.37 … Spain:… 7m      2017-0…
##  4 asv.6  SAMN1…   15401 SAMN10… SAMN10… filter… 43.34 … Spain:… surface 2017-0…
##  5 asv.5  SAMN1…   12944 SAMN10… SAMN10… settle… 43.37 … Spain:… 7m      2017-1…
##  6 asv.11 SAMN1…   12292 SAMN10… SAMN10… settle… 43.34 … Spain:… 7m      2017-0…
##  7 asv.14 SAMN1…   11142 SAMN10… SAMN10… filter… 43.37 … Spain:… surface 2017-0…
##  8 asv.10 SAMN1…   10447 SAMN10… SAMN10… filter… 43.37 … Spain:… surface 2017-0…
##  9 asv.3  SAMN1…   10377 SAMN10… SAMN10… settle… 43.34 … Spain:… 7m      2017-1…
## 10 asv.18 SAMN1…    9850 SAMN10… SAMN10… settle… 43.37 … Spain:… 7m      2017-0…
## # … with 117,184 more rows, 27 more variables: occurrenceStatus <chr>,
## #   target_gene <chr>, subfragment <chr>, pcr_primer_forward <chr>,
## #   pcr_primer_reverse <chr>, pcr_primer_name_forw <chr>,
## #   pcr_primer_name_reverse <chr>, pcr_primer_reference <chr>,
## #   lib_layout <chr>, seq_meth <chr>, sop <chr>, votu_db <chr>, My_name <chr>,
## #   kingdom <chr>, phylum <chr>, class <chr>, order <chr>, family <chr>,
## #   genus <chr>, species <chr>, lastvalue <chr>, otu_seq_comp_appr <chr>, …

Exploratory data analysis

stats_phyla <- df %>%
  group_by(phylum) %>%
  summarize(abundance = sum(Abundance)) %>%
  arrange(desc(abundance))

top_phyla <- head(na.omit(stats_phyla$phylum), 8)
top_phyla

## [1] "Arthropoda"      "Cnidaria"        "Bacillariophyta" "Annelida"       
## [5] "Bryozoa"         "Nemertea"        "Chlorophyta"     "Mollusca"

phylum_colors <- RColorBrewer::brewer.pal(9, "Spectral")
names(phylum_colors) <- c(top_phyla, "Other")
phylum_colors

##      Arthropoda        Cnidaria Bacillariophyta        Annelida         Bryozoa 
##       "#D53E4F"       "#F46D43"       "#FDAE61"       "#FEE08B"       "#FFFFBF" 
##        Nemertea     Chlorophyta        Mollusca           Other 
##       "#E6F598"       "#ABDDA4"       "#66C2A5"       "#3288BD"

df <- df %>%
  mutate(phylum_binned = ifelse(is.na(phylum) | phylum %in% top_phyla, phylum, "Other")) %>%
  mutate(phylum_binned = factor(phylum_binned, levels = c(top_phyla, "Other")))

# remotes::install_github("david-barnett/microViz")

physeq <- physeq %>%
  microViz::tax_mutate(phylum = ifelse(phylum == "", NA, phylum)) %>%
  microViz::tax_mutate(phylum_binned = ifelse(is.na(phylum) | phylum %in% top_phyla, phylum, "Other"))

library(tidyr)

stats_type_phylum <- df %>%
  group_by(eventRemarks, phylum) %>%
  summarize(abundance = sum(Abundance))

spread(stats_type_phylum, eventRemarks, abundance) %>%
  mutate(total = `settlement plates` + `filtered water`) %>%
  arrange(desc(total)) %>%
  knitr::kable()

library(ggplot2)

stats_type_phylum <- df %>%
  # calculate total abundance by sample type
  group_by(eventRemarks) %>%
  mutate(abundance = Abundance / sum(Abundance)) %>%
  group_by(eventRemarks, phylum_binned) %>%
  summarize(abundance = sum(abundance))
  
ggplot() +
  geom_bar(data = stats_type_phylum, aes(y = eventRemarks, fill = phylum_binned, x = abundance), stat = "identity") +
  scale_fill_manual(values = phylum_colors, na.value = "#eeeeee") +
  theme_minimal()

stats_sample_phylum <- df %>%
  # calculate total abundance by sample type
  group_by(Sample) %>%
  mutate(relative_abundance = Abundance / sum(Abundance)) %>%
  group_by(Sample, eventRemarks, phylum_binned) %>%
  summarize(relative_abundance = sum(relative_abundance), abundance = sum(Abundance))

ggplot() +
  geom_bar(data = stats_sample_phylum, aes(y = Sample, fill = phylum_binned, x = abundance), stat = "identity") +
  scale_fill_manual(values = phylum_colors, na.value = "#eeeeee") +
  theme_minimal() +
  theme(panel.grid.major.y = element_blank(), panel.grid.minor.y = element_blank())

ggplot() +
  geom_bar(data = stats_sample_phylum, aes(y = Sample, fill = phylum_binned, x = relative_abundance), stat = "identity") +
  scale_fill_manual(values = phylum_colors, na.value = "#eeeeee") +
  theme_minimal() +
  theme(panel.grid.major.y = element_blank(), panel.grid.minor.y = element_blank())

top_species <- df %>%
  filter(!is.na(species)) %>%
  group_by(species) %>%
  summarize(abundance = sum(Abundance)) %>%
  arrange(desc(abundance)) %>%
  head(20)

top_species %>% knitr::kable()

stats <- df %>%
  filter(species %in% top_species$species) %>%
  group_by(species, eventRemarks, Sample) %>%
  summarize(abundance = sum(Abundance))

ggplot() +
  geom_jitter(data = stats, aes(y = species, x = abundance, color = eventRemarks), pch = 21, height = 0.1, width = 0) +
  scale_color_brewer(palette = "Set1") +
  theme_minimal()

Invasive species

In this section we will check our results against the World Register of Introduced Marine Species (WRiMS). The repository contains a CSV file containing the LSIDs for all species in WRiMS.

wrims_lsids <- read.csv("wrims_lsids.csv")

invasives <- df %>%
  filter(lsid %in% wrims_lsids$lsid)

invasives %>%
  group_by(phylum, species) %>%
  summarize(abundance = sum(Abundance)) %>%
  arrange(desc(abundance)) %>%
  rmarkdown::paged_table()

ASV accumulation curves

To do.

Alpha and beta diversity

To do.

Multidimensional scaling with phyloseq

Phyloseq’s ordinate() function allows us to do MultiDimensional Scaling (MDS) or Principal Coordinate Analysis (PCoA) on our dataset. The default for this function is to use Bray-Curtis Dissimilarity to calculate a distance matrix.

ord <- ordinate(physeq = physeq, method = "PCoA", distance = "bray")

plot_ordination(physeq = physeq, ordination = ord, type = "samples", color = "eventRemarks", shape = "locality") +
  geom_point(aes(color = eventRemarks), alpha = 1, size = 4) +
  geom_text(aes(label = materialSampleID), alpha = 0.7, size = 3, vjust = 2) +
  stat_ellipse(aes(group = eventRemarks)) +
  scale_color_brewer(palette = "Set1") +
  scale_shape(solid = TRUE) +
  theme_classic()

We can also add taxa to the ordination plot, but styling this is a bit awkward as samples and taxa are treated as a single dataset.

plot_ordination(physeq = subset_taxa(physeq, !is.na(phylum)), ordination = ord, type = "biplot", color = "phylum_binned") +
  scale_color_manual(values = phylum_colors, na.value = "#000000") +
  geom_text(aes(label = materialSampleID), alpha = 0.7, size = 3, vjust = 2) +
  theme_classic()

Non-metric MDS (NMDS) with phyloseq

Another ordination method is non-metric MDS (NMDS). Let’s try NMDS with our dataset aggregated at the species level, and plot the most abundant species together with the sites.

First aggregates the phyloseq object to the species level:

physeq_species <- tax_glom(physeq, taxrank = "species", NArm = TRUE, bad_empty = c(NA, ""))

Then perform the ordination with method NMDS:

ord_species <- ordinate(physeq_species, method = "NMDS", distance = "bray")

Extract the NMDS values for the sites and species, and join with the relevant tables from the phyloseq object to get the necessary metadata for plotting (such as species name, locality, and sample type):

sites <- ord_species$points %>%
  as.data.frame() %>%
  tibble::rownames_to_column("site") %>%
  inner_join(physeq_species@sam_data %>% as.data.frame() %>% tibble::rownames_to_column("site"), by = "site")

species <- ord_species$species %>%
  as.data.frame() %>%
  tibble::rownames_to_column("asv") %>%
  left_join(physeq_species@tax_table %>% as.data.frame() %>% tibble::rownames_to_column("asv"), by = "asv") %>%
  filter(species %in% top_species$species)

ggplot() +
  geom_segment(data = species, aes(x = 0, y = 0, xend = MDS1, yend = MDS2), arrow = arrow(length = unit(2, "mm")), color = "#aaaaaa") +
  geom_point(data = sites, aes(MDS1, MDS2, color = eventRemarks, shape = locality), size = 3, stroke = 1.5) +
  geom_text(data = species, aes(MDS1, MDS2, label = species), size = 3, vjust = "outward") +
  scale_color_brewer(palette = "Set1") +
  scale_shape(solid = FALSE) +
  theme_classic()